Huge Radiators: Roundup

[HESmelaugh]

hey Shane, any problems with the QC with the XSPC rad? i have been reading a lot about missing fins, bent fins etc. i know that XSPC has solved the paint problem which came with the first shipment, but I would like to know if it has solved the bent and missing fins problem as well...

The RX480 I got is literally flawless. The fins are perfectly spaced, the finish is smooth and even, all the threads are in the right place etc.
If my RX480 represents all the current RX rads then there is really nothing to worry about. But of course I can't really know since I only have that one radiator here.

Shane it would of been nice to see the TFC 480 In there, As not many people test Them, Not sure if it because you can't get TFC via aqua tuning, where you work.

But it will be nice to see the comparison between this and martins, to see how different the results are.

Agreed on both points.
Just like with the 360s, I would obviously love to get my hands on a TFC 480. I've only heard good things about those rads and it would be great to see how they stack up agains the ones I've tested so far.

The problem is that not only don't I get samples from TFC, their products are actually available nowhere in Switzerland...
If they were, I'd just buy one. But if I have to buy one and pay for the rad, the international shipping and taxes (we pay lots of taxes here, for stuff that crosses the border), I'm paying ridiculous amounts of money for just a radiator...

Awesome work...man you have been busy!!

I'm suprised the Mo-Ra is so restrictive, I would have never guessed it would even be more restrictive than a GTS. It must have some very thin tubes or something causing that.

Anyhow, good to know in the 120 vs 140mm fans on the monsta, I suspected 140's are really needed to make full use of it's 140mm factor. Time to go order some, I think I'll go with the Koolance fans since they go up to 1700RPM.

Yeah, it's been a few busy weeks. ^^
I was also surprised at how restrictive the Mo-Ra is. Tubes are 9mm afaik. I flushed it with water, like all the other rads and I assume that was good enough. But I think I'll reflush it with some vinegar and see if there's any more gunk in there. If there is -> retest.
I was also surprised to see how much of a difference fan size made on the Monsta. Another argument for 140 mm fans, IMO.

you have been busy HES! I'm glad you got the C/W calculations licked, also nice to see the data table as well. Nice work!

Although, should this test be battle of the fatties versus "huge".

Thanks!
Though I have to say, I am still a bit sceptical towards my results. Looking at the data table, there are some weird numbers in there...
I don't think that my results are dramatically wrong, but I'm guessing they're still a bit off here and there.

can one of you pro's out there explain to me the whole basis behind the C/W calculations and why it is so widely used?
i've tried googling it... kept getting some damned TV station...

sorry for my nub question...

A radiators performance shows in it's temperature difference "Water - Air". i.e. how much hotter does the water get than the ambient temperature? The smaller this difference, the better the rad performs.

This difference is dependant on the radiators efficiency at transferring heat (in short: it's performance) and the amount of heat it has to deal with. The more heat is being dumped, the higher the temperature difference is going to be, obviously.

Now, C/W is exactly that measurement - temperature difference Water - Air - but adjusted to how much heat was dumped into the loop during testing. The heaters don't deliver a completely constant heat load, so just comparing the absolute temp differences isn't quite fair.

So if your data shows that radiator X produced a temp difference Water - Air of, say, 7.5° at a given fan speed, that's one part of the equation. Let's say that the average heatload during the test was 290W.

This results in a C/W of 7.5/290 = 0.025862

To translate C/W values back to tangible water temperatures, just multiply them with the estimated heatload in your LC loop and add your room temperature to that.

With the above example, I would say my system dumps maybe 180W of heat into the loop so it's 0.025862*180 = 4.66° plus my room temperature 24° = 28.66° water temperature.
So I could roughly estimate that I'd get a water temp of about 29° with a radiator that has a C/W of 0.025862

So much for my explanation. I hope it helps (and I also hope I didn't make any mistakes...).

[kinghong1970]

The RX480 I got is literally flawless. The fins are perfectly spaced, the finish is smooth and even, all the threads are in the right place etc.
If my RX480 represents all the current RX rads then there is really nothing to worry about. But of course I can't really know since I only have that one radiator here.

i was playing around with my RX480 and i realized that my has one flaw... the 4 corners of the radiator does not lay flat against a flat surface.
it's a bit "twisted" to give appx 1/8" wobble at 2 opposing corners.

meh... so long as it hold water and cools it...

and now... for this part...

A radiators performance shows in it's temperature difference "Water - Air". i.e. how much hotter does the water get than the ambient temperature? The smaller this difference, the better the rad performs.

This difference is dependant on the radiators efficiency at transferring heat (in short: it's performance) and the amount of heat it has to deal with. The more heat is being dumped, the higher the temperature difference is going to be, obviously.

Now, C/W is exactly that measurement - temperature difference Water - Air - but adjusted to how much heat was dumped into the loop during testing. The heaters don't deliver a completely constant heat load, so just comparing the absolute temp differences isn't quite fair.

So if your data shows that radiator X produced a temp difference Water - Air of, say, 7.5° at a given fan speed, that's one part of the equation. Let's say that the average heatload during the test was 290W.

This results in a C/W of 7.5/290 = 0.025862

ok, so according to your chart, you have temps for air in and air out as well as temps for water in and water out...

which figures do you use for the temp difference of "Water -Air" ?

To translate C/W values back to tangible water temperatures, just multiply them with the estimated heatload in your LC loop and add your room temperature to that.

as for estimation of your heatload, where and how?

With the above example, I would say my system dumps maybe 180W of heat into the loop so it's 0.025862*180 = 4.66° plus my room temperature 24° = 28.66° water temperature.
So I could roughly estimate that I'd get a water temp of about 29° with a radiator that has a C/W of 0.025862

So much for my explanation. I hope it helps (and I also hope I didn't make any mistakes...).

aaah, i get this part... you're basically taking the CW value and plugging it backwards to get appx water temps in your loop when exiting the radiator...

i think my brain cell just split!!!
cell count hittin double digits now!!!

watch out you 1337 testers!!! evolution is in action!

[Migi06]

I forgot to mention this: The Mo-Ra comes with spacers and screws for mounting it to the side-panel of a case. It's also wide enough to just stand on it's own.

Could Push&Pull done with Mora 2 core and with 2 fan faceplates?(not pro, because you cant mount another one so well)

Just out of curiosity: Where do you get that estimate from?
Generally, a 140 mm fan can be potentially more noisy than a 120 mm fan because it's pushing more air at identical rpm. Below a certain threshhold though, rpm makes no audiable difference in a quality fan, i.e.

I cant know exact number.. its more guess than fact, just like comparing 80mm fan to 92mm or 120mm fan (need to compare same manufacture fan). Bigger fan make more noise in same rpm because its produce a lot more airflow wich will make "hummm" effect.
80mm Nexus(19 dBA@1m: 1110 RPM :15 CFM) vs 120mm Nexus(22 dBA@1m: 1080 RPM: 29 CFM). Found in SPCR site.

I wouldn't be able to tell if it's doing 300rpm or 500rpm from the distance I'm usually away from my fans.

You are right 300rpm vs 500rpm is too hard to hear (because they are death silent)
Its like above 600rpm at 120mm fans when they come audible..

All I'm trying to say is: I guess the noise-level relations between fan formats are pretty complex.

Like you said there are many things what affect to noise.. Bearings, rpm, horizontal/vertical, wing type, fan count, yourself (everyone can't hear same noises) etc.. also sound can change comparing free air moving fan vs Radiator fan or insade case vs outside of the case..

[HESmelaugh]

ok, so according to your chart, you have temps for air in and air out as well as temps for water in and water out...

which figures do you use for the temp difference of "Water -Air" ?

Water = Average of water in and water out
Air = Air in

as for estimation of your heatload, where and how?

I have a watt-meter connected to the heaters and I simply write down the value it shows once every five minutes. This is a bit of a rough method, of course, but the fluctuations in this value are very gradual so I think I still get a fair average like this.

Could Push&Pull done with Mora 2 core and with 2 fan faceplates?(not pro, because you cant mount another one so well)

I'm sure it could be done somehow. For example, you could also connect all of the fans to each other, resulting in a "wall of fans" that you could then attach to a Mo-Ra on either side quite easily. Maybe some DIY mounting or even just cable-ties and duct-tape.

Concerning fan-noise: Thanks for the quote from spcr. I'd almost forgotten about that site.
Ultimately, I think the issue of noise is a very sensitive one, like you describe. So far, I favour leaving the noise factor completely out of my testing. Makes my life a lot easier.

[kinghong1970]

thank you very much for your patience in helping me understand this...

it's greatly appreciated...

[Martinm210]

Any chance you can convert these C/W charts over to Heat dissipated at 10C or 5C?

It's just 10/(C/W), and would probably be helpful for alot of folks.

[NaeKuh]

Each test run lasts 35 minutes, 25 minutes are warm up and the data from the last 10 minutes is all averaged out.

Hes i hate to do this to ya... but i dont think you had enough warmup time.. Expecially on those large rads...

Courtesy of martin:

The problem with delta T testing I see is how long it can take to stabilize the run at high deltas and high heat loads with large volumes. I'm seeing some tests taking over 90 minutes to stabilize.

So much for production testing..lol!....

Sorry bro, you need 90 min on average, or your just testing inital hold capacity and not disapation. You need a much longer warm up sequence...

Skinnee had simular results when dealing with full copper rads.. Your warm up time is just too small.

and as u guys can see its a logrithmic at the start so you cant do a straight liner eq.

[Waterlogged]

I currently don't have any rad twice anymore. I sold/gave away the extra ones from the sandwich-testing.

But I could maybe throw the RX480 and RX360 together in a loop. They have practically identical restriction, it seems.

EDIT: I'm pretty sure almost no one misunderstands the Mo-Ra, by the way. The reason it is popular with those who it is popular with is that it offers a tidy package with tons of cooling power at low fan speeds. If you aren't obsessed with flowrates and as-low-as-possible temperatures and are more focused on as-low-as-possible noise levels, a Mo-Ra is a great radiator. It's a lot less bother than installing three separate rads and it's also cheaper than three good rads.

Those 2 RX's would suffice for the testing I have in mind.

The thing is, you don't need 3 "good" rads to equal or best the MO-RA, all you'd really need is 2, at least...that's my thinking.

[HESmelaugh]

I pretested the necessary time-frame for repeatable results. For whatever reason, I'm not seeing anything similar to the graphs shown above.

I should mention though, that there are two types of temperature courses in my testing procedure: The first run starts with the lowest fan speeds and the water warms up from ambient. I pre-warm the water a little bit, even though it's probably not necessary. So i start the heater, shut it off again after a few minutes, start the logging software and start the heater again. This is a temperature log picked out and just plotted quickly in Excel, showing the entire 35 minutes of a test-run:




After that, the water never completely cools down between test runs, I unplug the heater, adjust fan speeds to the next higher step and then start the heating again. I guess I could also just leave the heater running. Anyway, this means that for all but the first test run, the water is cooling down from what it was at lower fan speeds to a new equilibrium at higher fan speeds. This is such a curve, also picked out of the logs. It shows the change from 400rpm to 800rpm:



So, as you can see, I really don't see this gradual temperature change over a long period of time. The temps seem to reach a certain level after just a few minutes and then just stick to that.

I'm pretty sure I posted the results from 35 min vs 60 min test runs in my setup somewhere in these forums. Can't find it anymore, though.


EDIT @Waterlogged: That might be true. Two rads like the RX might outdo or match a Mo-Ra.

[dsumanik]

Sorry bro, you need 90 min on average, or your just testing inital hold capacity and not disapation. You need a much longer warm up sequence...

I think what you have to do when looking at this data is simply eat it with a grain of salt from a consumer perspective, not a scientific one. Logically you can simply think, if I plug this rad/fan combo into any given loop, it affects temperatures as so.

Which is what his data gives us, how this rad/fan combo will affect a loop, not efficiency or exact performance. The consumer can then rationalize the cost vs.benefits.

The data as tested cannot define the actual performance winner of these rads because of the fan sizes (as he stated in the introduction of his post)

Specifically looking at the specs of the fans he selected:

140mm 1500 RPM 82.8 CFM
120mm 1900 RPM 75.0 CFM

At full voltage a 400 RPM defecit on the 140mm Has a 7.8 CFM advantage. I would wager that the higher RPMS give better static pressure on the 120mm.

So, how can we possibly compare these radiator based on RPM's?

Perhaps RPM should be thrown out of radiator testing when trying to discover actual dissipation for different sized rads. If all the rads are 120's, or 140's you can get away with it because the initial airflow is constant...but when you compare different formats the airflow needs to be controlled somehow, otherwise the results will skew. So if we calculate C/W with different sized rads against RPM, the data only gives us a trend as to how the radiator is best suited (hi or low speed fan), which is still very useful, but for actual dissipation the temps need to be plotted VS airflow in my opinion...at least if they are going to be on the same graph.

Maybe a single large fan directing a known quantity of air through a shroud with a fixed size opening (i.e a 100x100 square) would be a better way to test dissipation. The recorded temps would be a result of constant static pressure and airflow for a given surface area, call it potential dissipation/inch or something. You could then multiply by surface area to get a decent estimate. Leave the flowrates uncontrolled so this characteristic of the rad is factored in as well.

Another option... maybe in order to compare a 120mm rad vs 140mm rad you need to consider 120mm and 140mm fans with airflow specs that converge over an RPM range, this wouldn't provide similar static pressure, but at lease would eliminate one more variable.

maybe one day martin can do some 140mm fan testing and we can identify the best fans to mate with a 120 vs 140 rad comparison

[Spoiler]

I think what you have to do when looking at this data is simply eat it with a grain of salt from a consumer perspective, not a scientific one. Logically you can simply think, if I plug this rad/fan combo into any given loop, it affects temperatures as so.

Which is what his data gives us, how this rad/fan combo will affect a loop, not efficiency or exact performance. The consumer can then rationalize the cost vs.benefits.

The data as tested cannot define the actual performance winner of these rads because of the fan sizes (as he stated in the introduction of his post)

Specifically looking at the specs of the fans he selected:

140mm 1500 RPM 82.8 CFM
120mm 1900 RPM 75.0 CFM

At full voltage a 400 RPM defecit on the 140mm Has a 7.8 CFM advantage. I would wager that the higher RPMS give better static pressure on the 120mm.

So, how can we possibly compare these radiator based on RPM's?

Perhaps RPM should be thrown out of radiator testing when trying to discover actual dissipation for different sized rads. If all the rads are 120's, or 140's you can get away with it because the initial airflow is constant...but when you compare different formats the airflow needs to be controlled somehow, otherwise the results will skew. So if we calculate C/W with different sized rads against RPM, the data only gives us a trend as to how the radiator is best suited (hi or low speed fan), which is still very useful, but for actual dissipation the temps need to be plotted VS airflow in my opinion...at least if they are going to be on the same graph.

Maybe a single large fan directing a known quantity of air through a shroud with a fixed size opening (i.e a 100x100 square) would be a better way to test dissipation. The recorded temps would be a result of constant static pressure and airflow for a given surface area, call it potential dissipation/inch or something. You could then multiply by surface area to get a decent estimate. Leave the flowrates uncontrolled so this characteristic of the rad is factored in as well.

Another option... maybe in order to compare a 120mm rad vs 140mm rad you need to consider 120mm and 140mm fans with airflow specs that converge over an RPM range, this wouldn't provide similar static pressure, but at lease would eliminate one more variable.

maybe one day martin can do some 140mm fan testing and we can identify the best fans to mate with a 120 vs 140 rad comparison

I agree with you when saying you need to take these tests with a grain of salt. However, I look at these tests hoping that I would get an idea of how they might perform if I bought one. If those tests didn't reflect the actual heat dissipation due to limited testing duration, then the accuracy could be quite different.

[NaeKuh]

HES's graph looks good...

You must have a very small loop. And i mean a TINY loop. :P

But on my fesser's which are pure copper, i noticed the warnup time being a ...

[Snowman89]

Nice review! I have a gts 240 and it's not even considerd good and i must say im very happy with it. With 2fans running 1500ish rpm and a gtx 260 65nm and a phenom II 940 @ 3.4ghz my gpu is doing full load of 50ish in games, cpu about same temp.

[silverphoenix]

Awesome work...man you have been busy!!

I'm suprised the Mo-Ra is so restrictive, I would have never guessed it would even be more restrictive than a GTS. It must have some very thin tubes or something causing that.

Anyhow, good to know in the 120 vs 140mm fans on the monsta, I suspected 140's are really needed to make full use of it's 140mm factor. Time to go order some, I think I'll go with the Koolance fans since they go up to 1700RPM.

it's tubes are round but It is in essence a ton of passes which equates to alot of 180 degree turns. I'm absolutely surprised by the GTS420 in push pull.

[quattro_]

. I'm absolutely surprised by the GTS420 in push pull.

same here im thinking of getting one and use S-Flex F's push pull .

[RCG_Bex]

Nice review. I'd of like to have seen the new Koolance 480's thrown in there seeing as how they're a fair abit skinnier than most 480's I see.

I don't think I've ever actually read a review for my TFC480 either! But I agree with Naekuh, heatup times are longer than my Swiftech.

~Bex

[Martinm210]

Yeah, I think heat up times are primarily a factor of water volume in the loop and the delta between tests. My particular setup is huge, I have about a 1 gallon reservoir that serves to handle my heaters, then the large radiator, and I also have a large flow meter and a large gate valve, large tubing...etc. In the end I have over a full gallon in my test loop which has an incredible ability to store heat. Those inline heaters with little/no volume are a real benefit to reducing testing time.

[BringerOdeath]

I am thinking about about changing my GTX480 and GTX240 in to RX480 and GTX240.....What do you guys think? Or should i go for 240mm rads.....

[Martinm210]

This is a temperature log picked out and just plotted quickly in Excel, showing the entire 35 minutes of a test-run:

I thought the BigNG sensors were only good to .5C?

How often does it record/log (What does 339) represent?

And how the heck are you keeping your ambient that flat. No matter how hard I try, mine will slowly go up and down just a bit.

And one more stupid question of mine. How are you fixing RPM at exactly 400, 800, RPM etc.? Is that something the BigNG can do for you, if so..that's pretty handy. I've tried setting an RPM, but it varies slightly as the fans warm up and as system voltage fluctuates. I ended up giving up and just let the crystalfontz log the RPM every second and average the end result over the test.

The graph looks good, I can only wish mine would get stable that fast..

[HESmelaugh]

HES's graph looks good...

You must have a very small loop. And i mean a TINY loop. :P

But on my fesser's which are pure copper, i noticed the warnup time being a ...

That's actually an impotant point: I have my loop as short as possible. I also make sure that the EK X-Res on the pump is never filled to the top. Silly, but true. :p
So I have the absolute minimum of water in my loop.

I thought the BigNG sensors were only good to .5C?

How often does it record/log (What does 339) represent?

And how the heck are you keeping your ambient that flat. No matter how hard I try, mine will slowly go up and down just a bit.

And one more stupid question of mine. How are you fixing RPM at exactly 400, 800, RPM etc.? Is that something the BigNG can do for you, if so..that's pretty handy. I've tried setting an RPM, but it varies slightly as the fans warm up and as system voltage fluctuates. I ended up giving up and just let the crystalfontz log the RPM every second and average the end result over the test.

The graph looks good, I can only wish mine would get stable that fast..

Some very crucial points here, that aren't apparent just from plucking the data out of Excel.

Let me start with the fans: Unfortunately i can't fix fan speeds. They do fluctuate and I can't log the rpm. The reason for this is that the bigNG does not tell you the actual fan speeds at all. As far as I can tell, it simply checks the rpm once and then gives you calculated numbers on what the fan speeds should be at the voltage you set. This is why I actually use two Kaze Masters to monitor fan speeds. I warm the fans up for about ten minutes and then set them as closely to the target speed as I can i.e. fluctuating around the target speed. This is a source of impresicion in my testing. I'm sure I mentioned this in my first radiator review(s), but I should add this info to the testing method bit of the new review as well.

Water-Temp: The plot shows all four probes averaged out. It's correct that the bigNG has a resolution of .5°.

I made a new graph and put a little more effort into it, this time.
Here is the raw data from the water and air in probes:



It's logged in 5-sec-intervals. The air temperatures fluctuate by +/-.5 degrees for the entire range of the data that I average out for my results. I didn't include the air out temps because those are rubbish in my test setup...

[Martinm210]

That's actually an impotant point: I have my loop as short as possible. I also make sure that the EK X-Res on the pump is never filled to the top. Silly, but true. :p
So I have the absolute minimum of water in my loop.



Some very crucial points here, that aren't apparent just from plucking the data out of Excel.

Let me start with the fans: Unfortunately i can't fix fan speeds. They do fluctuate and I can't log the rpm. The reason for this is that the bigNG does not tell you the actual fan speeds at all. As far as I can tell, it simply checks the rpm once and then gives you calculated numbers on what the fan speeds should be at the voltage you set. This is why I actually use two Kaze Masters to monitor fan speeds. I warm the fans up for about ten minutes and then set them as closely to the target speed as I can i.e. fluctuating around the target speed. This is a source of impresicion in my testing. I'm sure I mentioned this in my first radiator review(s), but I should add this info to the testing method bit of the new review as well.

Water-Temp: The plot shows all four probes averaged out. It's correct that the bigNG has a resolution of .5°.

I made a new graph and put a little more effort into it, this time.
Here is the raw data from the water and air in probes:



It's logged in 5-sec-intervals. The air temperatures fluctuate by +/-.5 degrees for the entire range of the data that I average out for my results. I didn't include the air out temps because those are rubbish in my test setup...

Very nice!

Now that I go back and look at some recent runs, it does appear RPM is fairly stable, I think it all depends on the fan controller. In some of my earlier testing I was using a Sunbeam controller to set voltage, and RPM varied +- 30RPM or more, but I think it was the controller. In my recent testing, I've been using my variable PSU to set RPM and it's holding RPM spot on with 5RPM or so. I decide with this round to fix RPM at certain voltages, but it really doesn't matter how it's done if you're generating a trendline.

.5C isn't very good with just one sensor, but you've got 2 on the in and 2 on the out, so the 4 sensors total helps improve that resolution alot..

Awesome job on this. My only recommendation would be to compare either w/c which is heat dissipated per degree celcius, or compare something like a heat dissipate for a 10C delta which is probably more standard.

I find that comparing C/W doesn't appropriately scale the true performance, particularly at Higher RPM. It gives you the impression that the differences in higher RPM regions are very small and insignificant, but they are not.

Give it a try, you'll see what I mean. You have all the data, just copy another column in there and divide 10 by C/W

[Brodholm]

AFAIK, they only had problems with the early RX360. I never read about problems with the RX480.
Though it's been stated that they will switch to a matte finish and non-threaded fan holes - something I'm not terribly enthusiastic about.

OK, thanks. Why does they go with that new standard? Does it give any improvements or something? I know the thermochill have a black matt finish also. Could this have something to do with XSPC owning some of thermochill?

- Black Matt Finish
- Copper Tubes and Fins
- 3 rows of 13mm by 2mm Tubes
- Sub 8fpi (fins per inch)
- G1/4" Ports
- 6-32 Case and Fan Screws

Supplied with mounting screws

The first version of the RX radiator has a gloss black finish and M4 Screws, version 2 spec is listed above

[skinnee]

Actually, I think the screw change is from testers griping about the M4 screws.

[Brodholm]

Actually, I think the screw change is from testers griping about the M4 screws.

Well thats to bad, I like M4 :\

What about the color?

[HESmelaugh]

I had some time to tinker around with the data again and tried my hand at a "watts dissipated" graph:



What has me worried about this is that it scatters my data points all over the place.

It would be nice to get more precision here. I think the graphs above show that longer load times wouldn't help me here. High-precision thermal probes of some type would be very nice, but they won't do much good if I can't hook them up to the PC to have them log their readings somehow. Any suggestions?
I was considering a Crystalfontz a while back, btw. The reason I didn't go for it is because while it has higher resolution than .5°, it's thermal probes have a precision of only .5°. So it was either the bigNG (very precise digital probes, but only .5° resolution) or the Fontz (high reolution, .5° precision probes). Seems to me that neither is better than the other...

@posts above: What the hell is wrong with M4 screws? How could threadless holes possibly be better than M4?